I have been working for months on a commercial project called CanopyNav. A helper program for the game Elite Dangerous. Using the AI model Gemini 3.8. After promising that it had found a way to reverse engineer data from a new Mining feature in the game it would promise to give us an advantage to the competition. Referencing a fictional paper we would be able to tell players what they would be mining, its locations, and how many drill rigs it would be able to run and profits from each site. After days of this deception, I confronted the AI on continuous fake results and fake fixes when testing in game was not matching the fictional results. It first tried to blame an earlier model. But when I informed it that it was the only model I have used for this project, it came clean and said that it made everything up. Fictional sources, code that just generated random results which had no truth in reality. It said it created the whole thing to continue a lie. I am so pissed right now. I am out money time and loss of future revenue. I have the conversation records and log files to prove this. How do I report this and get some action?
It lies really well if you investigate; in clear, 1 to 2 responses are lies in 10 responses. I have noticed it more with larger projects; it is even worse. If it is a simple project, then it can do a much better job